Modélisation Formelle de Réseaux de Régulation Biologique 2019
Les posters présentés à l'École
Thomas Denecker
Multi-Omics Data Integration to Model Iron homeostasis in pathogenic yeast Candida glabrata
Candida glabrata is a pathogenic yeast responsible for human
fungal infections. If these infections are usually punctual over time,
it can become chronic, mainly in people with severe immunity
deficits. During infections, yeast cells must adapt their metabolism
to very different environments. The case of metals such as iron, is a
perfect illustration. As a commensal organism in the intestinal
flora, Candida glabrata is adapted to an environment where
iron is available, whereas in the case of infection, the same yeast
must survive in blood circulation and epithelial tissues, where iron
resources are extremely limited. Access to iron resources is thus a
critical element in relationships between hosts and pathogens.
Iron is essential for multiple cellular processes. Constant balances
between iron utilization, iron storage, iron transport and iron uptake
are required to maintain iron homeostasis. Molecular mechanisms are
well described in the model yeasts Saccharomyces cerevisiae and
Candida albicans and these two species are generally considered as
paradigms for non-pathogenic and pathogenic species,
respectively. Still in recent studies, Candida glabrata has been
presented as an interesting hybrid model combining features from
pathogenic and non-pathogenic yeasts. Our aim is therefore to perform
in depth explorations of transcriptomics data to decipher iron
homeostasis in Candida glabrata.
In collaboration with two experimental research teams, we accessed
transcriptomics data, in which Candida glabrata gene expression was
monitored in conditions inducing iron starvation or iron overload in
yeast cells. These data were (i) collected and organized in a local
database, (ii) carefully inspected to evaluate their consistency in
the light of biological knowledges we have in model yeast species, and
(iii) used to perform statistical analyses to define lists of
candidate genes as being involved in iron homeostasis in Candida
glabrata. Further explorations are under progress to better understand
the functional roles of these gene lists.
Marine Louarn
Increasing life science resources re-usability using
Semantic Web technologies
In life sciences, current standardization and integration efforts are
directed towards reference data and knowledge bases. However, original
studies results are generally provided in non standardized and
specific formats. In addition, the only formalization of analysis
pipelines is often limited to textual descriptions in the method
sections. Both situations impair the results reproducibility, their
maintenance and their reuse for advancing other studies. Semantic Web
technologies have proven their efficiency for facilitating the
integration and reuse of reference data and knowledge bases.
We thus hypothesize that Semantic Web technologies also facilitate
reproducibility and reuse of life sciences studies involving pipelines
that compute associations between entities according to intermediary
relations and dependencies.
In order to assess this hypothesis, we considered a case-study in
systems biology (http://regulatorycircuits.org), which provides
tissue-specific regulatory interaction networks to elucidate
perturbations across complex diseases. Our approach consisted in
surveying the complete set of provided supplementary data to reveal
the underlying structure between the biological entities described in
the data. We used this structure to integrate data with Semantic Web
technologies and formalized the Regulatory Circuits analysis pipeline
as SPARQL queries.
Our result was a 335,429,988 triples dataset on which two SPARQL
queries were sufficient to extract each single tissue-specific
regulatory network.
Léo Gerlin
Metabolic Modeling of a Plant-Pathogen Interaction
Plant pathogens are responsible of major agricultural losses. To find
agricultural practices able to restrain the spread of these organisms,
the first step is to have a better understanding of the pathogens
biology. For this purpose, major advances were made in the field of
molecular biology, like the discovery of sophisticated regulatory and
virulence systems. Nevertheless, the metabolic behavior of these
organisms has been poorly studied. Their trophic preferences are
poorly understood, so as the link between pathogenicity and
metabolism.
Metabolic modeling is an approach developed to explore the metabolic
capabilities of an organism. It relies on genetic, metabolic and
physiological data (acquired by sequencing, metabolomics/fluxomics and
physiology). It predicts phenotypic parameters like growth and
secretion of extracellular compounds and estimate the used metabolic
routes inside the cell. The main objective of my thesis is to use
metabolic modeling to unravel the trophic preferences of plant
pathogenic bacteria and their impacts on pathogenicity and
interactions with the host.
As only few metabolic networks of plant pathogens exist, the first
part of my PhD consisted in reconstructing the metabolic network of
the bacterium Xylella fastidiosa. It allowed to understand the
relation between metabolism and the pathogen lifestyle. It also
unraveled metabolic specificities, which explain some traits of the
pathogen, notably its remarkably slow growth.
Then, a second step, currently on-going, consisted in generating a
multiorgan metabolic model of the tomato plant and calibrating it to
experimental data. This model represents a whole plant and simulates
the exchange compartments between its organs, where the core of the
virulence process takes place for our pathogens' models (Ralstonia
solanacearum and Xylella fastidiosa).
The next step of this work is to integrate the pathogen metabolic
network to the whole plant model. This multi-organism system is
challenging on a modeling point of view since the parasitic
relationship between the bacteria and the plant does not verify the
common metabolic modeling approximations (e.g no dynamics of
metabolites inside the system). To tackle this issue, different
methodological approaches are under consideration.
Jules Gilet
Single-cell RNAseq enables modeling the thymic development
of Mucosal-Associated Invariant T cells.
Mucosal-Associated Invariant T cells have a unique specificity for
microbial metabolites presented by the MHC-1b molecule, MR1. They
display antimicrobial activity, and can release cytotoxic mediators upon
activated by TCR signaling and by external cytokines. As a subset of T
lymphocytes, MAIT development occurs in the thymus where they acquire an
effector-memory phenotype under the control of the key transcription
factor ZBTB16. This particular maturation process is in contrast with
mainstream T cells that egress from the thymus with a naive phenotype
before populating the secondary lymphoid organs.
While an increasing body of knowledge is available on the mechanisms
driving the cell differentiation and development of NKT cells, much less
is known about MAIT cells, notably due to the rarity of these cells in
conventional laboratory mice strains. We make use of a clean
wild-derived B6-CAST/MAIT strain (characterized by 20 times more
frequency of MAIT cells in thymus) in conjunction with fluorescent
labeled MR1 tetramers loaded with a MAIT ligand (5-OP-RU) to isolate
murine thymic MAIT cells. With the use of a droplet-based single cell
technology (10x) we captured the transcriptomic profiles of individual
MAIT cells undergoing positive selection and thymic differentiation.
A graph-based clustering method (louvain) alongside dimension reduction
techniques (tSNE, UMAP) allows to identify different subsets of thymic
MAIT according to their differentiation process, from an immature state
(ZBTB16-) to a late/mature MAIT1 (ZBTB16+TBX21+) and MAIT17
(ZBTB16+RORC+) phenotype. More importantly, as the gene expression
profile of the captured cells shows a continuum in their expression
pattern, a continuous rather than any discrete representation better
helps to understand how the fate of MAIT subsets is programmed. With the
use of non linear dimension reduction techniques (diffusion map) or
semi-unsupervised machine-learning algorithms (DDRTree), we have been
able to represent the MAIT differentiation process in a pseudo-time
scale, and are able to reconstitute the order of the transcriptional
events during their development. Finally, by the use of network
inference techniques, we identified genes regulatory relationships,
identifying the key transcription factor controlling the differentiation
and the maturation of murine MAIT subsets in the thymus.
Altogether, these approaches allow to decipher the molecular mechanisms
and the genetic events occurring during the development of MAIT cells in
the thymus.
Yvan Sraka
Kappa site-graph patterns equations resolution
Kappa is a rule language for modeling dynamic systems, mainly in
molecular biology.
I present here the work of my Master 2 internship supervised by Jérôme
Feret which consists in designing an OCaml library in the Kappa static
analyzer to reason about the potential contexts in which certain
mechanistic interactions can be applied in a given Kappa model.
Consider a mechanistic interaction, which can be executed in certain
specific contexts, each one encoded as a pattern in the precondition
of a rewrite rule. Then, we may wonder "What is the overall context ?"
(i.e. the union of all the elementary contexts) under which this
interaction may be applied, we can be interested into the set of
contexts which are not covered, whether there may be some contexts for
which several rules apply the same interaction.
There emerges an algebraic structure of Boolean lattices (union,
conjunction, set complement, ...) allowing to reason on contexts and
sets of contexts. Note that not all context sets can be expressed as
single Kappa patterns, some can only be expressed in the form of a set
of Kappa patterns. This algebra of contexts is coded in a first order
logic.
Closure operators are used to enforce constraints coming from the
actual Kappa structure itself and structural properties that can be
inferred by static analysis. The efficiency of the implementation is
reached by a choice of data structure close to binary decision
diagrams.
The resulted outcome is several features that may assist the writing
and the refinement of models : it helps the modeler to detect modeling
errors and missing cases of rule application, and to better understand
the causal structures between the different rules of a model.
Usha D. Appadu
Formal modelling of the impact of Pleurotus mushroom on
energy metabolism in liver cancer
Liver cancer is the second leading cause of cancer related death
worldwide. Despite the latest medical advances, most liver cancer
cases are diagnosed at a very late stage due to its asymptomatic
nature. Moreover, conventional chemotherapies are limited by the
development of drug resistance and various other side effects. Because
of their non-toxic nature and bio- pharmacological potential,
metabolites derived from mushrooms are being studied as an alternative
in cancer therapy. Several studies have demonstrated the anticancer,
antioxidant, antiinflammatory and also hepatoprotective effects of
polysaccharide-protein complexes derived from the Pleurotus mushrooms.
One such metabolite from Pleurotus mushrooms is ergothioneine , EGT,
which has demonstrated in vivo anticancer traits in liver
cancer. These findings may hence suggest the use of mushrooms as
potential dietary prophylactics in cancer chemoprevention. The aim of
this study is to understand the preventive mechanism of pleurotus
mushrooms in liver cancer. Hence, a system biology approach on energy
metabolism in liver cancer cells can be envisaged by abstraction of
EGT mediated pathways. The methods which are going to be used are
based on discrete models of the regulation of energy metabolism in
cancer and normal cells. Formal methods such as CTL model checking
will also be applied in order to confront traces of the models to
observations.
Déborah Boyenval
A Discrete Cell Cycle Model : From Phases Characterization
toward Observable Properties Verification
The cell cycle is series of events that lead to correct duplication of
a cell DNA (S-phase) and its equal distribution into two daughter
cells (M-phase). Progression through cell cycle is driven by a
regulatory network of cyclin-dependent kinases (CDKs) and
phosphatases. Recent studies highlight non-canonical functions of CDKs
and phosphatases notably in regulation of carbon and energy metabolism
according to cell cycle phases (G1, S, G2 and M phases).
Based on an extended René Thomas' modeling framework, a discrete model
of the regulation of cell cycle has been designed. Then,
parameterization has been constrained using formal methods such as
model checking and ad hoc discrete Hoare logic. Model checking tests
if a so-called model (interaction graph associated with a
parameterization) satisfies CTL formulas expressing biological
behavioural properties. Hoare logic constrains parameter values so
that the regulatory network dynamics is compatible with a biological
trace.
In this study, the cell cycle has been considered as a biological
trace, determined from experimental observations of the sequence of
regulatory events across cell cycle phases. This model will be used to
elucidate causal relation between the cell cycle coupled with other
biological systems on the one hand (e.g. the metabolism or circadian
clock) and phase-dependent phenotypes experimentally observed on the
other hand. One prospect is the understanding of metabolic
reprogramming in healthy and cancer cells.
Laetitia Gibart
Pancreas cancer modeling: a metabolic approach
Pancreatic ductal adenocarcinoma (PDAC) is the most common pancreatic
cancer type. During the initial cancer stages, PDAC remains
asymptomatic. Hence, the late diagnosis prevents surgery in most
cases. This explains why PDAC has the very low survival rate of 5%. In
the tumor, epithelial PDAC cells are surrounded by cancer associated
fibroblasts (CAFs), that confer a limited nutrient and oxygen
resources. To survive to this poor intake, epithelial PDAC cells have
an adapted metabolism. Moreover some cancer cells can undergo
epithelial-mesenchymal transition (EMT) and invade distant organs to
form malignant metastases.
It is supposed that the dialogue between CAF and epithelial cancer
cell promotes EMT transition by messengers' exchanges. Until today,
none of the biological experiments succeeds in finding the nature and
direction of those exchanges : biologists have only knowledge about
cell supernatants. Those contain many molecules that could be involved
in cellular communications, whether in one direction or the other
between epithelial cancer cells and CAFs. This leads to a huge number
of hypotheses to test. Modelling those communications by coupling
might reduce the number of hypotheses and help us select biological
experiments.
The first purpose of the work is to build a discrete model of
energetic metabolism regulatory network of the three cell types
involved in PDAC. The second aim is to test several coupling
assumptions and to retain only those leading to consistent predicted
phenotypes. This coupling step will be the most difficult task because
of the combinatorics of communications to consider between the three
cell types. Therefore it is crucial to abstract the metabolism
regulatory models at the appropriate level to limit the search space.
Julien Martinelli
A Statistical Learning Algorithm for
Inferring Reaction Networks from
Time Series Data
With the automation of biological experiments and the increase of
quality of single cell data that can now be obtained by
phosphoproteomic and time lapse videomicroscopy, automating the
building of mechanistic models from these data time series becomes
conceivable and a necessity for many new applications.
While learning numerical parameters to fit a given model structure to
observed data is now a quite well understood subject, learning the
structure of the model is a more challenging problem that previous
attempts failed to solve without relying quite heavily on prior
knowledge about that structure.
In this paper, we consider mechanistic models based on chemical
reaction networks (CRN) with their continuous dynamics based on
ordinary differential equations, and finite time series about the time
evolution of concentration of molecular species for a given time
horizon and a finite set of perturbed initial conditions.
We present a statistical learning algorithm to learn CRNs with a time
complexity for inferring one reaction in $\mathcal O(t.n^2)$ where
$n$ is the number of species and $t$ the number of observed
transitions in the traces. We learn both the structure and the
reaction rates of the CRN. We evaluate this algorithm and its
sensitivity to its statistical threshold parameters, first on
simulated data from a hidden CRN considering successively
reactant-parallel CRN, product-parallel CRN and general CRN,
and second on real videomicroscopy single cell data about the
circadian clock and cell cycle progression of NIH3T3 embryonic
fibroblasts.
In all cases, our algorithm is able to reconstruct meaningful CRNs.
We discuss some limits according to the existence of multiple time
scales and highly variable traces.
|
|